PART A¶

1. Data import and Understanding¶

A. Read the ‘Signals.csv’ as DataFrame and import required libraries

In [1]:
from google.colab import drive
drive.mount('/content/drive')

Mounted at /content/drive
In [17]:
import pandas as pd
import numpy as np
sd = pd.read_csv('/content/drive/MyDrive/AIML Course/Introduction to Neural Networks/Part-+1%2C2%263+-+Signal.csv')
In [18]:
sd.shape
Out[18]:
(1599, 12)
In [19]:
sd.head()
Out[19]:
Parameter 1 Parameter 2 Parameter 3 Parameter 4 Parameter 5 Parameter 6 Parameter 7 Parameter 8 Parameter 9 Parameter 10 Parameter 11 Signal_Strength
0 7.4 0.70 0.00 1.9 0.076 11.0 34.0 0.9978 3.51 0.56 9.4 5
1 7.8 0.88 0.00 2.6 0.098 25.0 67.0 0.9968 3.20 0.68 9.8 5
2 7.8 0.76 0.04 2.3 0.092 15.0 54.0 0.9970 3.26 0.65 9.8 5
3 11.2 0.28 0.56 1.9 0.075 17.0 60.0 0.9980 3.16 0.58 9.8 6
4 7.4 0.70 0.00 1.9 0.076 11.0 34.0 0.9978 3.51 0.56 9.4 5
In [20]:
sd.tail()
Out[20]:
Parameter 1 Parameter 2 Parameter 3 Parameter 4 Parameter 5 Parameter 6 Parameter 7 Parameter 8 Parameter 9 Parameter 10 Parameter 11 Signal_Strength
1594 6.2 0.600 0.08 2.0 0.090 32.0 44.0 0.99490 3.45 0.58 10.5 5
1595 5.9 0.550 0.10 2.2 0.062 39.0 51.0 0.99512 3.52 0.76 11.2 6
1596 6.3 0.510 0.13 2.3 0.076 29.0 40.0 0.99574 3.42 0.75 11.0 6
1597 5.9 0.645 0.12 2.0 0.075 32.0 44.0 0.99547 3.57 0.71 10.2 5
1598 6.0 0.310 0.47 3.6 0.067 18.0 42.0 0.99549 3.39 0.66 11.0 6
In [21]:
sd.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1599 entries, 0 to 1598
Data columns (total 12 columns):
 #   Column           Non-Null Count  Dtype  
---  ------           --------------  -----  
 0   Parameter 1      1599 non-null   float64
 1   Parameter 2      1599 non-null   float64
 2   Parameter 3      1599 non-null   float64
 3   Parameter 4      1599 non-null   float64
 4   Parameter 5      1599 non-null   float64
 5   Parameter 6      1599 non-null   float64
 6   Parameter 7      1599 non-null   float64
 7   Parameter 8      1599 non-null   float64
 8   Parameter 9      1599 non-null   float64
 9   Parameter 10     1599 non-null   float64
 10  Parameter 11     1599 non-null   float64
 11  Signal_Strength  1599 non-null   int64  
dtypes: float64(11), int64(1)
memory usage: 150.0 KB
In [11]:
import warnings
warnings.filterwarnings("ignore")
import numpy as np
from numpy.random import seed
seed(1)
import pandas as pd
import seaborn as sns
import scipy.stats as stats 
import matplotlib.pyplot as plt
import tensorflow as tf
import random
random.seed(1)
np.random.seed(1) 
tf.random.set_seed(1)
from tensorflow import keras
from keras.models import Sequential
from keras.layers import Dense
from sklearn.model_selection import StratifiedKFold
%matplotlib inline
#Test Train Split
from sklearn.model_selection import train_test_split
#Feature Scaling library
from tensorflow.keras.utils import to_categorical
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Activation, Dense
from tensorflow.keras import regularizers, optimizers
from tensorflow.keras import initializers
from sklearn.metrics import r2_score
from sklearn.preprocessing import LabelBinarizer
from tensorflow.keras.models import load_model
from tensorflow.keras.utils import custom_object_scope
from sklearn.preprocessing import MinMaxScaler
from sklearn.preprocessing import StandardScaler
from sklearn.metrics import classification_report
from sklearn.metrics import confusion_matrix
import random
from tensorflow.keras import backend

B. Check for missing values and print percentage for each attribute

In [23]:
round(sd.isnull().sum() / sd.isnull().count() * 100, 2)
Out[23]:
Parameter 1        0.0
Parameter 2        0.0
Parameter 3        0.0
Parameter 4        0.0
Parameter 5        0.0
Parameter 6        0.0
Parameter 7        0.0
Parameter 8        0.0
Parameter 9        0.0
Parameter 10       0.0
Parameter 11       0.0
Signal_Strength    0.0
dtype: float64
In [24]:
sd.isnull().values.any()
Out[24]:
False
In [45]:
#No missing values found in the dataset

C. Check for presence of duplicate records in the dataset and impute with appropriate method

In [25]:
sd_duplicates = sd.duplicated()
print('Number of duplicates:', sd_duplicates.sum())

Number of duplicates: 240
In [26]:
from scipy.stats import skew

# calculating skewness of each column
sd_sk = sd.skew()

# print the skewness of each column
print(sd_sk)

Parameter 1        0.982751
Parameter 2        0.671593
Parameter 3        0.318337
Parameter 4        4.540655
Parameter 5        5.680347
Parameter 6        1.250567
Parameter 7        1.515531
Parameter 8        0.071288
Parameter 9        0.193683
Parameter 10       2.428672
Parameter 11       0.860829
Signal_Strength    0.217802
dtype: float64
In [40]:
# As we find all the data are positively skewed, imputing the dataset with median to check if we could minimize duplicates
for column in sd.columns:
    sd[column].fillna(sd[column].mean(), inplace=True)
In [41]:
# calculating skewness of each column
sd_sk = sd.skew()

# print the skewness of each column
print(sd_sk)

Parameter 1        0.982751
Parameter 2        0.671593
Parameter 3        0.318337
Parameter 4        4.540655
Parameter 5        5.680347
Parameter 6        1.250567
Parameter 7        1.515531
Parameter 8        0.071288
Parameter 9        0.193683
Parameter 10       2.428672
Parameter 11       0.860829
Signal_Strength    0.217802
dtype: float64
In [42]:
#As we did not see any imporvement with median imputation, droppign the duplicates into a new dataframe
sd.duplicated().value_counts()
Out[42]:
False    1359
True      240
dtype: int64
In [43]:
sd_dedup=sd.drop_duplicates()
In [44]:
sd_dedup.value_counts()
Out[44]:
Parameter 1  Parameter 2  Parameter 3  Parameter 4  Parameter 5  Parameter 6  Parameter 7  Parameter 8  Parameter 9  Parameter 10  Parameter 11  Signal_Strength
4.6          0.52         0.15         2.1          0.054        8.0          65.0         0.99340      3.90         0.56          13.1          4                  1
8.8          0.24         0.54         2.5          0.083        25.0         57.0         0.99830      3.39         0.54          9.2           5                  1
             0.41         0.64         2.2          0.093        9.0          42.0         0.99860      3.54         0.66          10.5          5                  1
             0.40         0.40         2.2          0.079        19.0         52.0         0.99800      3.44         0.64          9.2           5                  1
             0.37         0.48         2.1          0.097        39.0         145.0        0.99750      3.04         1.03          9.3           5                  1
                                                                                                                                                                   ..
7.4          0.36         0.30         1.8          0.074        17.0         24.0         0.99419      3.24         0.70          11.4          8                  1
                          0.29         2.6          0.087        26.0         72.0         0.99645      3.39         0.68          11.0          5                  1
             0.35         0.33         2.4          0.068        9.0          26.0         0.99470      3.36         0.60          11.9          6                  1
             0.29         0.38         1.7          0.062        9.0          30.0         0.99680      3.41         0.53          9.5           6                  1
15.9         0.36         0.65         7.5          0.096        22.0         71.0         0.99760      2.98         0.84          14.9          5                  1
Length: 1359, dtype: int64

D. Visualise distribution of the target variable.¶

In [46]:
sd_dedup.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 1359 entries, 0 to 1598
Data columns (total 12 columns):
 #   Column           Non-Null Count  Dtype  
---  ------           --------------  -----  
 0   Parameter 1      1359 non-null   float64
 1   Parameter 2      1359 non-null   float64
 2   Parameter 3      1359 non-null   float64
 3   Parameter 4      1359 non-null   float64
 4   Parameter 5      1359 non-null   float64
 5   Parameter 6      1359 non-null   float64
 6   Parameter 7      1359 non-null   float64
 7   Parameter 8      1359 non-null   float64
 8   Parameter 9      1359 non-null   float64
 9   Parameter 10     1359 non-null   float64
 10  Parameter 11     1359 non-null   float64
 11  Signal_Strength  1359 non-null   int64  
dtypes: float64(11), int64(1)
memory usage: 138.0 KB
In [47]:
# y is tharget variable created for signal strengt in the dataframe
y = sd_dedup['Signal_Strength']

# Creating a histogram of the target variable
sns.histplot(y, kde=False, bins=10)
plt.title('Histogram of Target Variable')
plt.xlabel('Target Variable')
plt.ylabel('Count')
plt.show()
In [48]:
# Create a density plot of the target variable
sns.kdeplot(y)
plt.title('Density Plot of Target Variable')
plt.xlabel('Target Variable')
plt.ylabel('Density')
plt.show()
In [53]:
#Signal strength 5 and 6 has been showing the best results of all the 11 parameters measured
In [52]:
# Create a pair plot of the predictor variables and the target variable
x=sd_dedup.iloc[:,:-1]
sns.pairplot(pd.concat([x, y], axis=1))
Out[52]:
<seaborn.axisgrid.PairGrid at 0x7f5fc458d0a0>
In [54]:
#Parameter 1 & 8 are positively correlated
#Parameter 1 & 9 are negatively correlated

E. Share insights from the initial data analysis (at least 2).

In [56]:
# Compute the correlation matrix
sd_dedup_corr = sd_dedup.corr()

# Set the figure size
plt.subplots(figsize=(12, 8))

# Visualize the correlation matrix as a heatmap
sns.heatmap(sd_dedup_corr, cmap='RdBu', annot=True)
plt.title('Correlation Heatmap')
plt.show()
In [57]:
#Parameter 9 is negatively skewed to parameter 1 with highest value of 0.69
#Paramter 3 & 8 are positively skewed to Parameter 1
#Parameter 6 is also positively skewed to Parameter 7
# Both the above said positive correlation have highest values of 0.67

2. Data preprocessing¶

A. Split the data into X & Y.

In [58]:
x
Out[58]:
Parameter 1 Parameter 2 Parameter 3 Parameter 4 Parameter 5 Parameter 6 Parameter 7 Parameter 8 Parameter 9 Parameter 10 Parameter 11
0 7.4 0.700 0.00 1.9 0.076 11.0 34.0 0.99780 3.51 0.56 9.4
1 7.8 0.880 0.00 2.6 0.098 25.0 67.0 0.99680 3.20 0.68 9.8
2 7.8 0.760 0.04 2.3 0.092 15.0 54.0 0.99700 3.26 0.65 9.8
3 11.2 0.280 0.56 1.9 0.075 17.0 60.0 0.99800 3.16 0.58 9.8
5 7.4 0.660 0.00 1.8 0.075 13.0 40.0 0.99780 3.51 0.56 9.4
... ... ... ... ... ... ... ... ... ... ... ...
1593 6.8 0.620 0.08 1.9 0.068 28.0 38.0 0.99651 3.42 0.82 9.5
1594 6.2 0.600 0.08 2.0 0.090 32.0 44.0 0.99490 3.45 0.58 10.5
1595 5.9 0.550 0.10 2.2 0.062 39.0 51.0 0.99512 3.52 0.76 11.2
1597 5.9 0.645 0.12 2.0 0.075 32.0 44.0 0.99547 3.57 0.71 10.2
1598 6.0 0.310 0.47 3.6 0.067 18.0 42.0 0.99549 3.39 0.66 11.0

1359 rows × 11 columns

In [138]:
y.value_counts()
Out[138]:
5    577
6    535
7    167
4     53
8     17
3     10
Name: Signal_Strength, dtype: int64
In [60]:
x.shape
Out[60]:
(1359, 11)
In [62]:
y.shape
Out[62]:
(1359,)
In [139]:
# Converting y to catergorical
yc=to_categorical(y,num_classes=12)

B. Split the data into train & test with 70:30 proportion.

In [140]:
from sklearn.model_selection import train_test_split

# X represents the feature variables, y represents the target variable
x_train, x_test, y_train, y_test = train_test_split(x, yc, test_size=0.3, random_state=42)

C. Print shape of all the 4 variables and verify if train and test data is in sync.

In [141]:
# Print the shape of the four variables
print("x_train shape:", x_train.shape)
print("x_test shape:", x_test.shape)
print("y_train shape:", y_train.shape)
print("y_test shape:", y_test.shape)

# Verify if train and test data are in sync
assert x_train.shape[0] == y_train.shape[0], "Error: x_train and y_train have different number of samples"
assert x_test.shape[0] == y_test.shape[0], "Error: x_test and y_test have different number of samples"

x_train shape: (951, 11)
x_test shape: (408, 11)
y_train shape: (951, 12)
y_test shape: (408, 12)

D. Normalise the train and test data with appropriate method

In [142]:
x_train.min()
Out[142]:
Parameter 1     4.60000
Parameter 2     0.16000
Parameter 3     0.00000
Parameter 4     0.90000
Parameter 5     0.01200
Parameter 6     2.00000
Parameter 7     6.00000
Parameter 8     0.99007
Parameter 9     2.86000
Parameter 10    0.33000
Parameter 11    8.40000
dtype: float64
In [143]:
x_train.max()
Out[143]:
Parameter 1      15.9000
Parameter 2       1.5800
Parameter 3       0.7800
Parameter 4      13.9000
Parameter 5       0.6110
Parameter 6      72.0000
Parameter 7     289.0000
Parameter 8       1.0032
Parameter 9       3.9000
Parameter 10      1.9800
Parameter 11     14.9000
dtype: float64
In [148]:
from sklearn.preprocessing import MinMaxScaler

# Create a scaler object
sd_scaler = MinMaxScaler()

# Fit the scaler object on the training data
sd_scaler.fit(x_train)

# Transform the training and testing data using the scaler object
x_train_norm = sd_scaler.transform(x_train)
x_test_norm = sd_scaler.transform(x_test)
In [149]:
x_train_norm.min()
Out[149]:
0.0
In [150]:
x_train_norm.max()
Out[150]:
1.0000000000000002
In [151]:
x_test_norm.min()
Out[151]:
-0.11538461538461497
In [152]:
x_test_norm.max()
Out[152]:
1.282051282051282

E. Transform Labels into format acceptable by Neural Network

In [153]:
print(y_train.shape)
print(y_test.shape)

(951, 12)
(408, 12)

3. Model Training & Evaluation using Neural Network¶

A. Design a Neural Network to train a classifier¶

In [163]:
sd_cls_model = Sequential()
sd_cls_model.add(Dense(8, activation='tanh'))
sd_cls_model.add(Dense(10, activation='tanh'))
sd_cls_model.add(Dense(12, activation='sigmoid'))
In [164]:
# Compile the model
sd_cls_model.compile(loss="categorical_crossentropy", metrics=["accuracy"], optimizer="Adam")

B. Train the classifier using previously designed Architecture¶

In [165]:
# Fit the model
history=sd_cls_model.fit(x_train, y_train, batch_size=20, epochs=50, validation_data=(x_test, y_test))

Epoch 1/50
48/48 [==============================] - 1s 7ms/step - loss: 2.3254 - accuracy: 0.0431 - val_loss: 1.9616 - val_accuracy: 0.4510
Epoch 2/50
48/48 [==============================] - 0s 4ms/step - loss: 1.7661 - accuracy: 0.4595 - val_loss: 1.5556 - val_accuracy: 0.4828
Epoch 3/50
48/48 [==============================] - 0s 4ms/step - loss: 1.4755 - accuracy: 0.4911 - val_loss: 1.3650 - val_accuracy: 0.5000
Epoch 4/50
48/48 [==============================] - 0s 4ms/step - loss: 1.3436 - accuracy: 0.4974 - val_loss: 1.2802 - val_accuracy: 0.4926
Epoch 5/50
48/48 [==============================] - 0s 3ms/step - loss: 1.2830 - accuracy: 0.4932 - val_loss: 1.2310 - val_accuracy: 0.5123
Epoch 6/50
48/48 [==============================] - 0s 3ms/step - loss: 1.2483 - accuracy: 0.5005 - val_loss: 1.2067 - val_accuracy: 0.5074
Epoch 7/50
48/48 [==============================] - 0s 4ms/step - loss: 1.2281 - accuracy: 0.4911 - val_loss: 1.1957 - val_accuracy: 0.4828
Epoch 8/50
48/48 [==============================] - 0s 4ms/step - loss: 1.2134 - accuracy: 0.4953 - val_loss: 1.1772 - val_accuracy: 0.5123
Epoch 9/50
48/48 [==============================] - 0s 3ms/step - loss: 1.2021 - accuracy: 0.4932 - val_loss: 1.1668 - val_accuracy: 0.5147
Epoch 10/50
48/48 [==============================] - 0s 3ms/step - loss: 1.1944 - accuracy: 0.4953 - val_loss: 1.1629 - val_accuracy: 0.4951
Epoch 11/50
48/48 [==============================] - 0s 3ms/step - loss: 1.1875 - accuracy: 0.4942 - val_loss: 1.1556 - val_accuracy: 0.5074
Epoch 12/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1819 - accuracy: 0.4984 - val_loss: 1.1494 - val_accuracy: 0.5098
Epoch 13/50
48/48 [==============================] - 0s 3ms/step - loss: 1.1766 - accuracy: 0.4995 - val_loss: 1.1491 - val_accuracy: 0.4951
Epoch 14/50
48/48 [==============================] - 0s 3ms/step - loss: 1.1718 - accuracy: 0.5058 - val_loss: 1.1441 - val_accuracy: 0.5025
Epoch 15/50
48/48 [==============================] - 0s 3ms/step - loss: 1.1701 - accuracy: 0.4963 - val_loss: 1.1388 - val_accuracy: 0.5147
Epoch 16/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1659 - accuracy: 0.5005 - val_loss: 1.1400 - val_accuracy: 0.5025
Epoch 17/50
48/48 [==============================] - 0s 3ms/step - loss: 1.1635 - accuracy: 0.5047 - val_loss: 1.1382 - val_accuracy: 0.5025
Epoch 18/50
48/48 [==============================] - 0s 3ms/step - loss: 1.1613 - accuracy: 0.5026 - val_loss: 1.1371 - val_accuracy: 0.5049
Epoch 19/50
48/48 [==============================] - 0s 3ms/step - loss: 1.1587 - accuracy: 0.5047 - val_loss: 1.1397 - val_accuracy: 0.4902
Epoch 20/50
48/48 [==============================] - 0s 3ms/step - loss: 1.1568 - accuracy: 0.5026 - val_loss: 1.1337 - val_accuracy: 0.5049
Epoch 21/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1550 - accuracy: 0.5068 - val_loss: 1.1316 - val_accuracy: 0.5074
Epoch 22/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1528 - accuracy: 0.5058 - val_loss: 1.1317 - val_accuracy: 0.5049
Epoch 23/50
48/48 [==============================] - 0s 3ms/step - loss: 1.1515 - accuracy: 0.5068 - val_loss: 1.1319 - val_accuracy: 0.5025
Epoch 24/50
48/48 [==============================] - 0s 3ms/step - loss: 1.1499 - accuracy: 0.5037 - val_loss: 1.1307 - val_accuracy: 0.4877
Epoch 25/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1483 - accuracy: 0.5110 - val_loss: 1.1251 - val_accuracy: 0.5098
Epoch 26/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1488 - accuracy: 0.5058 - val_loss: 1.1286 - val_accuracy: 0.4853
Epoch 27/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1464 - accuracy: 0.5142 - val_loss: 1.1270 - val_accuracy: 0.5147
Epoch 28/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1435 - accuracy: 0.5037 - val_loss: 1.1324 - val_accuracy: 0.4853
Epoch 29/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1452 - accuracy: 0.5100 - val_loss: 1.1281 - val_accuracy: 0.4779
Epoch 30/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1430 - accuracy: 0.5131 - val_loss: 1.1268 - val_accuracy: 0.4902
Epoch 31/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1403 - accuracy: 0.5163 - val_loss: 1.1210 - val_accuracy: 0.5025
Epoch 32/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1393 - accuracy: 0.5121 - val_loss: 1.1245 - val_accuracy: 0.4877
Epoch 33/50
48/48 [==============================] - 0s 3ms/step - loss: 1.1403 - accuracy: 0.5089 - val_loss: 1.1250 - val_accuracy: 0.4877
Epoch 34/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1384 - accuracy: 0.5110 - val_loss: 1.1173 - val_accuracy: 0.5098
Epoch 35/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1374 - accuracy: 0.5226 - val_loss: 1.1288 - val_accuracy: 0.4804
Epoch 36/50
48/48 [==============================] - 0s 3ms/step - loss: 1.1363 - accuracy: 0.5131 - val_loss: 1.1208 - val_accuracy: 0.5025
Epoch 37/50
48/48 [==============================] - 0s 3ms/step - loss: 1.1350 - accuracy: 0.5152 - val_loss: 1.1238 - val_accuracy: 0.4926
Epoch 38/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1347 - accuracy: 0.5121 - val_loss: 1.1161 - val_accuracy: 0.5123
Epoch 39/50
48/48 [==============================] - 0s 3ms/step - loss: 1.1333 - accuracy: 0.5100 - val_loss: 1.1150 - val_accuracy: 0.5123
Epoch 40/50
48/48 [==============================] - 0s 3ms/step - loss: 1.1319 - accuracy: 0.5247 - val_loss: 1.1199 - val_accuracy: 0.5025
Epoch 41/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1306 - accuracy: 0.5184 - val_loss: 1.1158 - val_accuracy: 0.5098
Epoch 42/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1290 - accuracy: 0.5226 - val_loss: 1.1149 - val_accuracy: 0.5098
Epoch 43/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1277 - accuracy: 0.5152 - val_loss: 1.1143 - val_accuracy: 0.5123
Epoch 44/50
48/48 [==============================] - 0s 3ms/step - loss: 1.1283 - accuracy: 0.5195 - val_loss: 1.1181 - val_accuracy: 0.4902
Epoch 45/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1269 - accuracy: 0.5163 - val_loss: 1.1102 - val_accuracy: 0.5098
Epoch 46/50
48/48 [==============================] - 0s 5ms/step - loss: 1.1254 - accuracy: 0.5342 - val_loss: 1.1190 - val_accuracy: 0.4853
Epoch 47/50
48/48 [==============================] - 0s 5ms/step - loss: 1.1264 - accuracy: 0.5226 - val_loss: 1.1172 - val_accuracy: 0.4951
Epoch 48/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1259 - accuracy: 0.5237 - val_loss: 1.1139 - val_accuracy: 0.4926
Epoch 49/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1224 - accuracy: 0.5384 - val_loss: 1.1142 - val_accuracy: 0.4926
Epoch 50/50
48/48 [==============================] - 0s 4ms/step - loss: 1.1216 - accuracy: 0.5258 - val_loss: 1.1088 - val_accuracy: 0.5147

C. Plot 2 separate visuals.¶

i. Training Loss and Validation Loss

ii. Training Accuracy and Validation Accuracy

In [166]:
# Plot the training and validation loss
plt.plot(history.history['loss'])
plt.plot(history.history['val_loss'])
plt.title('Training Loss vs. Validation Loss')
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.legend(['Training Loss', 'Validation Loss'])
plt.show()
In [167]:
# Plot the training and validation accuracy
plt.plot(history.history['accuracy'])
plt.plot(history.history['val_accuracy'])
plt.title('Training Accuracy vs. Validation Accuracy')
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.legend(['Training Accuracy', 'Validation Accuracy'])
plt.show()

D. Design new architecture/update existing architecture in attempt to improve the performance of the model¶

In [184]:
#Adding batch normalization and dropout after each dense layer except eh output layer to imporve performance of the model
from tensorflow.keras.layers import Dense, Dropout, BatchNormalization

sd_cls_model_v1 = Sequential()
sd_cls_model_v1.add(Dense(128, activation='relu'))
sd_cls_model_v1.add(BatchNormalization())
sd_cls_model_v1.add(Dropout(0.2))
sd_cls_model_v1.add(Dense(64, activation='relu'))
sd_cls_model_v1.add(BatchNormalization())
sd_cls_model_v1.add(Dropout(0.2))
sd_cls_model_v1.add(Dense(32, activation='relu'))
sd_cls_model_v1.add(BatchNormalization())
sd_cls_model_v1.add(Dropout(0.2))
sd_cls_model_v1.add(Dense(16, activation='relu'))
sd_cls_model_v1.add(BatchNormalization())
sd_cls_model_v1.add(Dropout(0.2))
sd_cls_model_v1.add(Dense(12, activation='softmax'))
# Compile the model
sd_cls_model_v1.compile(loss="categorical_crossentropy", metrics=["accuracy"], optimizer="sgd")
In [185]:
# Fit the model
history_v1=sd_cls_model_v1.fit(x_train, y_train, batch_size=24, epochs=100, validation_data=(x_test, y_test))

Epoch 1/100
40/40 [==============================] - 2s 12ms/step - loss: 2.9542 - accuracy: 0.0715 - val_loss: 2.4001 - val_accuracy: 0.0196
Epoch 2/100
40/40 [==============================] - 0s 5ms/step - loss: 2.5906 - accuracy: 0.1272 - val_loss: 2.1936 - val_accuracy: 0.3775
Epoch 3/100
40/40 [==============================] - 0s 5ms/step - loss: 2.3475 - accuracy: 0.2156 - val_loss: 2.0636 - val_accuracy: 0.4583
Epoch 4/100
40/40 [==============================] - 0s 4ms/step - loss: 2.0819 - accuracy: 0.3239 - val_loss: 1.8929 - val_accuracy: 0.4828
Epoch 5/100
40/40 [==============================] - 0s 5ms/step - loss: 1.9669 - accuracy: 0.3586 - val_loss: 1.6899 - val_accuracy: 0.5147
Epoch 6/100
40/40 [==============================] - 0s 5ms/step - loss: 1.8525 - accuracy: 0.4206 - val_loss: 1.6652 - val_accuracy: 0.5196
Epoch 7/100
40/40 [==============================] - 0s 5ms/step - loss: 1.7433 - accuracy: 0.4479 - val_loss: 1.5482 - val_accuracy: 0.5098
Epoch 8/100
40/40 [==============================] - 0s 6ms/step - loss: 1.6325 - accuracy: 0.4448 - val_loss: 1.4497 - val_accuracy: 0.5221
Epoch 9/100
40/40 [==============================] - 0s 5ms/step - loss: 1.5478 - accuracy: 0.4606 - val_loss: 1.4078 - val_accuracy: 0.4926
Epoch 10/100
40/40 [==============================] - 0s 4ms/step - loss: 1.4911 - accuracy: 0.4932 - val_loss: 1.3965 - val_accuracy: 0.4926
Epoch 11/100
40/40 [==============================] - 0s 5ms/step - loss: 1.4315 - accuracy: 0.4953 - val_loss: 1.4401 - val_accuracy: 0.4902
Epoch 12/100
40/40 [==============================] - 0s 5ms/step - loss: 1.4193 - accuracy: 0.4648 - val_loss: 1.3582 - val_accuracy: 0.5074
Epoch 13/100
40/40 [==============================] - 0s 5ms/step - loss: 1.3744 - accuracy: 0.4890 - val_loss: 1.2668 - val_accuracy: 0.5074
Epoch 14/100
40/40 [==============================] - 0s 5ms/step - loss: 1.3239 - accuracy: 0.4942 - val_loss: 1.2858 - val_accuracy: 0.4902
Epoch 15/100
40/40 [==============================] - 0s 5ms/step - loss: 1.3467 - accuracy: 0.5005 - val_loss: 1.3083 - val_accuracy: 0.5000
Epoch 16/100
40/40 [==============================] - 0s 5ms/step - loss: 1.3191 - accuracy: 0.4911 - val_loss: 1.3005 - val_accuracy: 0.5074
Epoch 17/100
40/40 [==============================] - 0s 5ms/step - loss: 1.3247 - accuracy: 0.4911 - val_loss: 1.2372 - val_accuracy: 0.4853
Epoch 18/100
40/40 [==============================] - 0s 5ms/step - loss: 1.2796 - accuracy: 0.4795 - val_loss: 1.2122 - val_accuracy: 0.5098
Epoch 19/100
40/40 [==============================] - 0s 5ms/step - loss: 1.2749 - accuracy: 0.5026 - val_loss: 1.1874 - val_accuracy: 0.5245
Epoch 20/100
40/40 [==============================] - 0s 4ms/step - loss: 1.2599 - accuracy: 0.5226 - val_loss: 1.2069 - val_accuracy: 0.4926
Epoch 21/100
40/40 [==============================] - 0s 4ms/step - loss: 1.2318 - accuracy: 0.5216 - val_loss: 1.2615 - val_accuracy: 0.5025
Epoch 22/100
40/40 [==============================] - 0s 4ms/step - loss: 1.2465 - accuracy: 0.5068 - val_loss: 1.1652 - val_accuracy: 0.5221
Epoch 23/100
40/40 [==============================] - 0s 4ms/step - loss: 1.2584 - accuracy: 0.5195 - val_loss: 1.1725 - val_accuracy: 0.5466
Epoch 24/100
40/40 [==============================] - 0s 5ms/step - loss: 1.2153 - accuracy: 0.5237 - val_loss: 1.1956 - val_accuracy: 0.5147
Epoch 25/100
40/40 [==============================] - 0s 4ms/step - loss: 1.2063 - accuracy: 0.5184 - val_loss: 1.1510 - val_accuracy: 0.5392
Epoch 26/100
40/40 [==============================] - 0s 4ms/step - loss: 1.2177 - accuracy: 0.5216 - val_loss: 1.1598 - val_accuracy: 0.5172
Epoch 27/100
40/40 [==============================] - 0s 5ms/step - loss: 1.2218 - accuracy: 0.5300 - val_loss: 1.1565 - val_accuracy: 0.5049
Epoch 28/100
40/40 [==============================] - 0s 6ms/step - loss: 1.1929 - accuracy: 0.5195 - val_loss: 1.2180 - val_accuracy: 0.5000
Epoch 29/100
40/40 [==============================] - 0s 7ms/step - loss: 1.2043 - accuracy: 0.5184 - val_loss: 1.1328 - val_accuracy: 0.5196
Epoch 30/100
40/40 [==============================] - 0s 7ms/step - loss: 1.1699 - accuracy: 0.5342 - val_loss: 1.1893 - val_accuracy: 0.5147
Epoch 31/100
40/40 [==============================] - 0s 6ms/step - loss: 1.1917 - accuracy: 0.5163 - val_loss: 1.1102 - val_accuracy: 0.5196
Epoch 32/100
40/40 [==============================] - 0s 7ms/step - loss: 1.1602 - accuracy: 0.5152 - val_loss: 1.1265 - val_accuracy: 0.5196
Epoch 33/100
40/40 [==============================] - 0s 5ms/step - loss: 1.1859 - accuracy: 0.5216 - val_loss: 1.1217 - val_accuracy: 0.5490
Epoch 34/100
40/40 [==============================] - 0s 7ms/step - loss: 1.2005 - accuracy: 0.5100 - val_loss: 1.1059 - val_accuracy: 0.5172
Epoch 35/100
40/40 [==============================] - 0s 7ms/step - loss: 1.1654 - accuracy: 0.5289 - val_loss: 1.1183 - val_accuracy: 0.5147
Epoch 36/100
40/40 [==============================] - 0s 6ms/step - loss: 1.1543 - accuracy: 0.5489 - val_loss: 1.0972 - val_accuracy: 0.5221
Epoch 37/100
40/40 [==============================] - 0s 6ms/step - loss: 1.1396 - accuracy: 0.5310 - val_loss: 1.2657 - val_accuracy: 0.4461
Epoch 38/100
40/40 [==============================] - 0s 6ms/step - loss: 1.1633 - accuracy: 0.5300 - val_loss: 1.0875 - val_accuracy: 0.5686
Epoch 39/100
40/40 [==============================] - 0s 6ms/step - loss: 1.1507 - accuracy: 0.5394 - val_loss: 1.0893 - val_accuracy: 0.5613
Epoch 40/100
40/40 [==============================] - 0s 6ms/step - loss: 1.1573 - accuracy: 0.5342 - val_loss: 1.0785 - val_accuracy: 0.5686
Epoch 41/100
40/40 [==============================] - 0s 7ms/step - loss: 1.1507 - accuracy: 0.5363 - val_loss: 1.1163 - val_accuracy: 0.5515
Epoch 42/100
40/40 [==============================] - 0s 6ms/step - loss: 1.1759 - accuracy: 0.5279 - val_loss: 1.0873 - val_accuracy: 0.5490
Epoch 43/100
40/40 [==============================] - 0s 7ms/step - loss: 1.1176 - accuracy: 0.5447 - val_loss: 1.0968 - val_accuracy: 0.5417
Epoch 44/100
40/40 [==============================] - 0s 5ms/step - loss: 1.1290 - accuracy: 0.5384 - val_loss: 1.2083 - val_accuracy: 0.4730
Epoch 45/100
40/40 [==============================] - 0s 4ms/step - loss: 1.1450 - accuracy: 0.5289 - val_loss: 1.1470 - val_accuracy: 0.5123
Epoch 46/100
40/40 [==============================] - 0s 5ms/step - loss: 1.1288 - accuracy: 0.5468 - val_loss: 1.1660 - val_accuracy: 0.4853
Epoch 47/100
40/40 [==============================] - 0s 4ms/step - loss: 1.1407 - accuracy: 0.5405 - val_loss: 1.2065 - val_accuracy: 0.4608
Epoch 48/100
40/40 [==============================] - 0s 4ms/step - loss: 1.1166 - accuracy: 0.5426 - val_loss: 1.0894 - val_accuracy: 0.5294
Epoch 49/100
40/40 [==============================] - 0s 5ms/step - loss: 1.1349 - accuracy: 0.5394 - val_loss: 1.1433 - val_accuracy: 0.4926
Epoch 50/100
40/40 [==============================] - 0s 5ms/step - loss: 1.1296 - accuracy: 0.5363 - val_loss: 1.0694 - val_accuracy: 0.5441
Epoch 51/100
40/40 [==============================] - 0s 5ms/step - loss: 1.1345 - accuracy: 0.5363 - val_loss: 1.0658 - val_accuracy: 0.5686
Epoch 52/100
40/40 [==============================] - 0s 4ms/step - loss: 1.1208 - accuracy: 0.5563 - val_loss: 1.1554 - val_accuracy: 0.4779
Epoch 53/100
40/40 [==============================] - 0s 5ms/step - loss: 1.1334 - accuracy: 0.5394 - val_loss: 1.0845 - val_accuracy: 0.5441
Epoch 54/100
40/40 [==============================] - 0s 5ms/step - loss: 1.1161 - accuracy: 0.5373 - val_loss: 1.0601 - val_accuracy: 0.5686
Epoch 55/100
40/40 [==============================] - 0s 5ms/step - loss: 1.1241 - accuracy: 0.5384 - val_loss: 1.0786 - val_accuracy: 0.5490
Epoch 56/100
40/40 [==============================] - 0s 4ms/step - loss: 1.1193 - accuracy: 0.5573 - val_loss: 1.0497 - val_accuracy: 0.5711
Epoch 57/100
40/40 [==============================] - 0s 4ms/step - loss: 1.1066 - accuracy: 0.5689 - val_loss: 1.0461 - val_accuracy: 0.5711
Epoch 58/100
40/40 [==============================] - 0s 4ms/step - loss: 1.1172 - accuracy: 0.5478 - val_loss: 1.0425 - val_accuracy: 0.5662
Epoch 59/100
40/40 [==============================] - 0s 4ms/step - loss: 1.1161 - accuracy: 0.5415 - val_loss: 1.1074 - val_accuracy: 0.5319
Epoch 60/100
40/40 [==============================] - 0s 5ms/step - loss: 1.1165 - accuracy: 0.5584 - val_loss: 1.0585 - val_accuracy: 0.5858
Epoch 61/100
40/40 [==============================] - 0s 5ms/step - loss: 1.1177 - accuracy: 0.5258 - val_loss: 1.0598 - val_accuracy: 0.5760
Epoch 62/100
40/40 [==============================] - 0s 5ms/step - loss: 1.1120 - accuracy: 0.5405 - val_loss: 1.0505 - val_accuracy: 0.5637
Epoch 63/100
40/40 [==============================] - 0s 5ms/step - loss: 1.1045 - accuracy: 0.5468 - val_loss: 1.0793 - val_accuracy: 0.5343
Epoch 64/100
40/40 [==============================] - 0s 4ms/step - loss: 1.1281 - accuracy: 0.5405 - val_loss: 1.0777 - val_accuracy: 0.5417
Epoch 65/100
40/40 [==============================] - 0s 5ms/step - loss: 1.1155 - accuracy: 0.5426 - val_loss: 1.0554 - val_accuracy: 0.5686
Epoch 66/100
40/40 [==============================] - 0s 4ms/step - loss: 1.1182 - accuracy: 0.5499 - val_loss: 1.1976 - val_accuracy: 0.4436
Epoch 67/100
40/40 [==============================] - 0s 4ms/step - loss: 1.1245 - accuracy: 0.5489 - val_loss: 1.3740 - val_accuracy: 0.4044
Epoch 68/100
40/40 [==============================] - 0s 6ms/step - loss: 1.1217 - accuracy: 0.5447 - val_loss: 1.0479 - val_accuracy: 0.5735
Epoch 69/100
40/40 [==============================] - 0s 5ms/step - loss: 1.1244 - accuracy: 0.5331 - val_loss: 1.0954 - val_accuracy: 0.5417
Epoch 70/100
40/40 [==============================] - 0s 5ms/step - loss: 1.0942 - accuracy: 0.5647 - val_loss: 1.0595 - val_accuracy: 0.5417
Epoch 71/100
40/40 [==============================] - 0s 4ms/step - loss: 1.0997 - accuracy: 0.5384 - val_loss: 1.0679 - val_accuracy: 0.5515
Epoch 72/100
40/40 [==============================] - 0s 4ms/step - loss: 1.1031 - accuracy: 0.5447 - val_loss: 1.1082 - val_accuracy: 0.4975
Epoch 73/100
40/40 [==============================] - 0s 4ms/step - loss: 1.1018 - accuracy: 0.5447 - val_loss: 1.0365 - val_accuracy: 0.5735
Epoch 74/100
40/40 [==============================] - 0s 4ms/step - loss: 1.1002 - accuracy: 0.5415 - val_loss: 1.0398 - val_accuracy: 0.5760
Epoch 75/100
40/40 [==============================] - 0s 4ms/step - loss: 1.0845 - accuracy: 0.5573 - val_loss: 1.0703 - val_accuracy: 0.5490
Epoch 76/100
40/40 [==============================] - 0s 5ms/step - loss: 1.0940 - accuracy: 0.5394 - val_loss: 1.0619 - val_accuracy: 0.5613
Epoch 77/100
40/40 [==============================] - 0s 4ms/step - loss: 1.1023 - accuracy: 0.5289 - val_loss: 1.0835 - val_accuracy: 0.5441
Epoch 78/100
40/40 [==============================] - 0s 4ms/step - loss: 1.0976 - accuracy: 0.5331 - val_loss: 1.0516 - val_accuracy: 0.5515
Epoch 79/100
40/40 [==============================] - 0s 5ms/step - loss: 1.0961 - accuracy: 0.5563 - val_loss: 1.0463 - val_accuracy: 0.5588
Epoch 80/100
40/40 [==============================] - 0s 5ms/step - loss: 1.1126 - accuracy: 0.5447 - val_loss: 1.1274 - val_accuracy: 0.5221
Epoch 81/100
40/40 [==============================] - 0s 6ms/step - loss: 1.0923 - accuracy: 0.5478 - val_loss: 1.0543 - val_accuracy: 0.5858
Epoch 82/100
40/40 [==============================] - 0s 5ms/step - loss: 1.0944 - accuracy: 0.5510 - val_loss: 1.0537 - val_accuracy: 0.5662
Epoch 83/100
40/40 [==============================] - 0s 4ms/step - loss: 1.1075 - accuracy: 0.5478 - val_loss: 1.0722 - val_accuracy: 0.5392
Epoch 84/100
40/40 [==============================] - 0s 5ms/step - loss: 1.1194 - accuracy: 0.5279 - val_loss: 1.0438 - val_accuracy: 0.5662
Epoch 85/100
40/40 [==============================] - 0s 5ms/step - loss: 1.1101 - accuracy: 0.5657 - val_loss: 1.1457 - val_accuracy: 0.4804
Epoch 86/100
40/40 [==============================] - 0s 6ms/step - loss: 1.0824 - accuracy: 0.5647 - val_loss: 1.0469 - val_accuracy: 0.5343
Epoch 87/100
40/40 [==============================] - 0s 4ms/step - loss: 1.0796 - accuracy: 0.5605 - val_loss: 1.0402 - val_accuracy: 0.5711
Epoch 88/100
40/40 [==============================] - 0s 4ms/step - loss: 1.1009 - accuracy: 0.5373 - val_loss: 1.0278 - val_accuracy: 0.5711
Epoch 89/100
40/40 [==============================] - 0s 4ms/step - loss: 1.1024 - accuracy: 0.5594 - val_loss: 1.2484 - val_accuracy: 0.4485
Epoch 90/100
40/40 [==============================] - 0s 5ms/step - loss: 1.0958 - accuracy: 0.5342 - val_loss: 1.0767 - val_accuracy: 0.5539
Epoch 91/100
40/40 [==============================] - 0s 5ms/step - loss: 1.0849 - accuracy: 0.5499 - val_loss: 1.1669 - val_accuracy: 0.4681
Epoch 92/100
40/40 [==============================] - 0s 5ms/step - loss: 1.0935 - accuracy: 0.5521 - val_loss: 1.0313 - val_accuracy: 0.5760
Epoch 93/100
40/40 [==============================] - 0s 4ms/step - loss: 1.0888 - accuracy: 0.5615 - val_loss: 1.0340 - val_accuracy: 0.5564
Epoch 94/100
40/40 [==============================] - 0s 4ms/step - loss: 1.0949 - accuracy: 0.5363 - val_loss: 1.0664 - val_accuracy: 0.5588
Epoch 95/100
40/40 [==============================] - 0s 6ms/step - loss: 1.0811 - accuracy: 0.5436 - val_loss: 1.0184 - val_accuracy: 0.5662
Epoch 96/100
40/40 [==============================] - 0s 7ms/step - loss: 1.0971 - accuracy: 0.5342 - val_loss: 1.0322 - val_accuracy: 0.5784
Epoch 97/100
40/40 [==============================] - 0s 6ms/step - loss: 1.0825 - accuracy: 0.5478 - val_loss: 1.0394 - val_accuracy: 0.5809
Epoch 98/100
40/40 [==============================] - 0s 6ms/step - loss: 1.0797 - accuracy: 0.5384 - val_loss: 1.0524 - val_accuracy: 0.5760
Epoch 99/100
40/40 [==============================] - 0s 6ms/step - loss: 1.0878 - accuracy: 0.5384 - val_loss: 1.0404 - val_accuracy: 0.5833
Epoch 100/100
40/40 [==============================] - 0s 7ms/step - loss: 1.0869 - accuracy: 0.5626 - val_loss: 1.1001 - val_accuracy: 0.5123
In [187]:
# Plot the training and validation loss
plt.plot(history_v1.history['loss'])
plt.plot(history_v1.history['val_loss'])
plt.title('Training Loss vs. Validation Loss')
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.legend(['Training Loss', 'Validation Loss'])
plt.show()
In [188]:
# Plot the training and validation accuracy
plt.plot(history_v1.history['accuracy'])
plt.plot(history_v1.history['val_accuracy'])
plt.title('Training Accuracy vs. Validation Accuracy')
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.legend(['Training Accuracy', 'Validation Accuracy'])
plt.show()
In [223]:
#The above model provided a slightly better accuracy.
#However, adding more dense layers, with less filters to improve accuracy
sd_cls_model_v2 = Sequential()

sd_cls_model_v2.add(Dense(32, activation='relu'))
sd_cls_model_v2.add(BatchNormalization())
sd_cls_model_v2.add(Dropout(0.2))
sd_cls_model_v2.add(Dense(28, activation='relu'))
sd_cls_model_v2.add(BatchNormalization())
sd_cls_model_v2.add(Dropout(0.2))
sd_cls_model_v2.add(Dense(24, activation='relu'))
sd_cls_model_v2.add(BatchNormalization())
sd_cls_model_v2.add(Dropout(0.2))
sd_cls_model_v2.add(Dense(12, activation='sigmoid'))
# Compile the model
sd_cls_model_v2.compile(loss="categorical_crossentropy", metrics=["accuracy"], optimizer="Adam")

Adding more epochs from 100 to 250 to imporve accuracy further

In [224]:
# Fit the model
history_v2=sd_cls_model_v2.fit(x_train, y_train, batch_size=24, epochs=250, validation_data=(x_test, y_test))

Epoch 1/250
40/40 [==============================] - 17s 11ms/step - loss: 2.9911 - accuracy: 0.0736 - val_loss: 2.5458 - val_accuracy: 0.0466
Epoch 2/250
40/40 [==============================] - 0s 4ms/step - loss: 2.5882 - accuracy: 0.1598 - val_loss: 2.1544 - val_accuracy: 0.4804
Epoch 3/250
40/40 [==============================] - 0s 5ms/step - loss: 2.2925 - accuracy: 0.2671 - val_loss: 1.9656 - val_accuracy: 0.5147
Epoch 4/250
40/40 [==============================] - 0s 5ms/step - loss: 2.1127 - accuracy: 0.3596 - val_loss: 1.8172 - val_accuracy: 0.5539
Epoch 5/250
40/40 [==============================] - 0s 5ms/step - loss: 1.9774 - accuracy: 0.3996 - val_loss: 1.6926 - val_accuracy: 0.5221
Epoch 6/250
40/40 [==============================] - 0s 5ms/step - loss: 1.8435 - accuracy: 0.4385 - val_loss: 1.5853 - val_accuracy: 0.5049
Epoch 7/250
40/40 [==============================] - 0s 5ms/step - loss: 1.6772 - accuracy: 0.4763 - val_loss: 1.4850 - val_accuracy: 0.5098
Epoch 8/250
40/40 [==============================] - 0s 4ms/step - loss: 1.6126 - accuracy: 0.4805 - val_loss: 1.4003 - val_accuracy: 0.4975
Epoch 9/250
40/40 [==============================] - 0s 5ms/step - loss: 1.4607 - accuracy: 0.5205 - val_loss: 1.3065 - val_accuracy: 0.5172
Epoch 10/250
40/40 [==============================] - 0s 5ms/step - loss: 1.4628 - accuracy: 0.4700 - val_loss: 1.2555 - val_accuracy: 0.5098
Epoch 11/250
40/40 [==============================] - 0s 5ms/step - loss: 1.3842 - accuracy: 0.4900 - val_loss: 1.2165 - val_accuracy: 0.5074
Epoch 12/250
40/40 [==============================] - 0s 5ms/step - loss: 1.3508 - accuracy: 0.4690 - val_loss: 1.1826 - val_accuracy: 0.5196
Epoch 13/250
40/40 [==============================] - 0s 4ms/step - loss: 1.2810 - accuracy: 0.5079 - val_loss: 1.1540 - val_accuracy: 0.5466
Epoch 14/250
40/40 [==============================] - 0s 5ms/step - loss: 1.2583 - accuracy: 0.4995 - val_loss: 1.1347 - val_accuracy: 0.5466
Epoch 15/250
40/40 [==============================] - 0s 4ms/step - loss: 1.2173 - accuracy: 0.5152 - val_loss: 1.1077 - val_accuracy: 0.5564
Epoch 16/250
40/40 [==============================] - 0s 5ms/step - loss: 1.1767 - accuracy: 0.5321 - val_loss: 1.1056 - val_accuracy: 0.5392
Epoch 17/250
40/40 [==============================] - 0s 5ms/step - loss: 1.1835 - accuracy: 0.5089 - val_loss: 1.1012 - val_accuracy: 0.5490
Epoch 18/250
40/40 [==============================] - 0s 7ms/step - loss: 1.1556 - accuracy: 0.5279 - val_loss: 1.1182 - val_accuracy: 0.5392
Epoch 19/250
40/40 [==============================] - 0s 7ms/step - loss: 1.1423 - accuracy: 0.5289 - val_loss: 1.0790 - val_accuracy: 0.5539
Epoch 20/250
40/40 [==============================] - 0s 7ms/step - loss: 1.1467 - accuracy: 0.5331 - val_loss: 1.0748 - val_accuracy: 0.5637
Epoch 21/250
40/40 [==============================] - 0s 7ms/step - loss: 1.1186 - accuracy: 0.5426 - val_loss: 1.0688 - val_accuracy: 0.5637
Epoch 22/250
40/40 [==============================] - 0s 7ms/step - loss: 1.1071 - accuracy: 0.5447 - val_loss: 1.0777 - val_accuracy: 0.5662
Epoch 23/250
40/40 [==============================] - 0s 6ms/step - loss: 1.1511 - accuracy: 0.5394 - val_loss: 1.0769 - val_accuracy: 0.5515
Epoch 24/250
40/40 [==============================] - 0s 7ms/step - loss: 1.0703 - accuracy: 0.5563 - val_loss: 1.0676 - val_accuracy: 0.5662
Epoch 25/250
40/40 [==============================] - 0s 6ms/step - loss: 1.0989 - accuracy: 0.5563 - val_loss: 1.0764 - val_accuracy: 0.5564
Epoch 26/250
40/40 [==============================] - 0s 7ms/step - loss: 1.1001 - accuracy: 0.5636 - val_loss: 1.0552 - val_accuracy: 0.5588
Epoch 27/250
40/40 [==============================] - 0s 6ms/step - loss: 1.0999 - accuracy: 0.5405 - val_loss: 1.0628 - val_accuracy: 0.5441
Epoch 28/250
40/40 [==============================] - 0s 7ms/step - loss: 1.0876 - accuracy: 0.5363 - val_loss: 1.0397 - val_accuracy: 0.5809
Epoch 29/250
40/40 [==============================] - 0s 7ms/step - loss: 1.0883 - accuracy: 0.5447 - val_loss: 1.0371 - val_accuracy: 0.5711
Epoch 30/250
40/40 [==============================] - 0s 6ms/step - loss: 1.0779 - accuracy: 0.5195 - val_loss: 1.0189 - val_accuracy: 0.5833
Epoch 31/250
40/40 [==============================] - 0s 6ms/step - loss: 1.0605 - accuracy: 0.5647 - val_loss: 1.0179 - val_accuracy: 0.5637
Epoch 32/250
40/40 [==============================] - 0s 7ms/step - loss: 1.0716 - accuracy: 0.5542 - val_loss: 1.0096 - val_accuracy: 0.5809
Epoch 33/250
40/40 [==============================] - 0s 6ms/step - loss: 1.0782 - accuracy: 0.5594 - val_loss: 1.0228 - val_accuracy: 0.5735
Epoch 34/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0850 - accuracy: 0.5384 - val_loss: 1.0097 - val_accuracy: 0.5882
Epoch 35/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0614 - accuracy: 0.5699 - val_loss: 1.0088 - val_accuracy: 0.5686
Epoch 36/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0704 - accuracy: 0.5521 - val_loss: 1.0191 - val_accuracy: 0.5784
Epoch 37/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0791 - accuracy: 0.5426 - val_loss: 1.0082 - val_accuracy: 0.5882
Epoch 38/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0614 - accuracy: 0.5489 - val_loss: 1.0088 - val_accuracy: 0.5735
Epoch 39/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0521 - accuracy: 0.5552 - val_loss: 1.0020 - val_accuracy: 0.5858
Epoch 40/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0562 - accuracy: 0.5605 - val_loss: 1.0092 - val_accuracy: 0.5833
Epoch 41/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0547 - accuracy: 0.5531 - val_loss: 0.9988 - val_accuracy: 0.5784
Epoch 42/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0508 - accuracy: 0.5584 - val_loss: 0.9983 - val_accuracy: 0.5833
Epoch 43/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0230 - accuracy: 0.5689 - val_loss: 0.9931 - val_accuracy: 0.5809
Epoch 44/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0500 - accuracy: 0.5468 - val_loss: 0.9981 - val_accuracy: 0.5784
Epoch 45/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0582 - accuracy: 0.5605 - val_loss: 0.9959 - val_accuracy: 0.5784
Epoch 46/250
40/40 [==============================] - 0s 6ms/step - loss: 1.0495 - accuracy: 0.5531 - val_loss: 1.0029 - val_accuracy: 0.5809
Epoch 47/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0439 - accuracy: 0.5415 - val_loss: 1.0131 - val_accuracy: 0.5833
Epoch 48/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0503 - accuracy: 0.5668 - val_loss: 1.0037 - val_accuracy: 0.5613
Epoch 49/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0458 - accuracy: 0.5521 - val_loss: 0.9981 - val_accuracy: 0.5833
Epoch 50/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0517 - accuracy: 0.5552 - val_loss: 0.9947 - val_accuracy: 0.5833
Epoch 51/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0339 - accuracy: 0.5531 - val_loss: 0.9979 - val_accuracy: 0.5858
Epoch 52/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0415 - accuracy: 0.5594 - val_loss: 0.9819 - val_accuracy: 0.5833
Epoch 53/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0345 - accuracy: 0.5626 - val_loss: 0.9756 - val_accuracy: 0.5833
Epoch 54/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0039 - accuracy: 0.5857 - val_loss: 0.9842 - val_accuracy: 0.5980
Epoch 55/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0280 - accuracy: 0.5584 - val_loss: 0.9675 - val_accuracy: 0.5784
Epoch 56/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0106 - accuracy: 0.5794 - val_loss: 0.9683 - val_accuracy: 0.5907
Epoch 57/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0191 - accuracy: 0.5794 - val_loss: 0.9685 - val_accuracy: 0.5833
Epoch 58/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0180 - accuracy: 0.5426 - val_loss: 0.9655 - val_accuracy: 0.5956
Epoch 59/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0369 - accuracy: 0.5384 - val_loss: 0.9755 - val_accuracy: 0.6005
Epoch 60/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0128 - accuracy: 0.5668 - val_loss: 0.9707 - val_accuracy: 0.5907
Epoch 61/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0258 - accuracy: 0.5741 - val_loss: 0.9767 - val_accuracy: 0.5931
Epoch 62/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0221 - accuracy: 0.5720 - val_loss: 0.9712 - val_accuracy: 0.5907
Epoch 63/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0152 - accuracy: 0.5668 - val_loss: 0.9816 - val_accuracy: 0.5907
Epoch 64/250
40/40 [==============================] - 0s 6ms/step - loss: 1.0286 - accuracy: 0.5636 - val_loss: 0.9812 - val_accuracy: 0.5907
Epoch 65/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0215 - accuracy: 0.5510 - val_loss: 0.9750 - val_accuracy: 0.5833
Epoch 66/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0221 - accuracy: 0.5762 - val_loss: 0.9815 - val_accuracy: 0.5858
Epoch 67/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0262 - accuracy: 0.5668 - val_loss: 0.9765 - val_accuracy: 0.5907
Epoch 68/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0255 - accuracy: 0.5521 - val_loss: 0.9743 - val_accuracy: 0.5784
Epoch 69/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0162 - accuracy: 0.5531 - val_loss: 0.9909 - val_accuracy: 0.5882
Epoch 70/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0059 - accuracy: 0.5668 - val_loss: 0.9745 - val_accuracy: 0.5882
Epoch 71/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0057 - accuracy: 0.5710 - val_loss: 0.9763 - val_accuracy: 0.5931
Epoch 72/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0069 - accuracy: 0.5626 - val_loss: 0.9751 - val_accuracy: 0.5833
Epoch 73/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0047 - accuracy: 0.5762 - val_loss: 0.9634 - val_accuracy: 0.5907
Epoch 74/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0128 - accuracy: 0.5584 - val_loss: 0.9640 - val_accuracy: 0.5809
Epoch 75/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0108 - accuracy: 0.5605 - val_loss: 0.9637 - val_accuracy: 0.5833
Epoch 76/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0164 - accuracy: 0.5563 - val_loss: 0.9705 - val_accuracy: 0.5931
Epoch 77/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0281 - accuracy: 0.5584 - val_loss: 0.9928 - val_accuracy: 0.5882
Epoch 78/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0208 - accuracy: 0.5573 - val_loss: 0.9712 - val_accuracy: 0.6078
Epoch 79/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9971 - accuracy: 0.5710 - val_loss: 0.9910 - val_accuracy: 0.5931
Epoch 80/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0470 - accuracy: 0.5468 - val_loss: 0.9741 - val_accuracy: 0.6054
Epoch 81/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0313 - accuracy: 0.5605 - val_loss: 0.9653 - val_accuracy: 0.6127
Epoch 82/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0084 - accuracy: 0.5584 - val_loss: 0.9836 - val_accuracy: 0.5980
Epoch 83/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0108 - accuracy: 0.5521 - val_loss: 0.9725 - val_accuracy: 0.5980
Epoch 84/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0134 - accuracy: 0.5657 - val_loss: 0.9580 - val_accuracy: 0.5980
Epoch 85/250
40/40 [==============================] - 0s 7ms/step - loss: 1.0218 - accuracy: 0.5657 - val_loss: 0.9900 - val_accuracy: 0.5907
Epoch 86/250
40/40 [==============================] - 0s 7ms/step - loss: 1.0049 - accuracy: 0.5710 - val_loss: 0.9743 - val_accuracy: 0.5858
Epoch 87/250
40/40 [==============================] - 0s 8ms/step - loss: 0.9949 - accuracy: 0.5710 - val_loss: 0.9797 - val_accuracy: 0.6152
Epoch 88/250
40/40 [==============================] - 0s 8ms/step - loss: 1.0077 - accuracy: 0.5468 - val_loss: 0.9726 - val_accuracy: 0.6152
Epoch 89/250
40/40 [==============================] - 0s 6ms/step - loss: 1.0073 - accuracy: 0.5552 - val_loss: 0.9901 - val_accuracy: 0.6152
Epoch 90/250
40/40 [==============================] - 0s 7ms/step - loss: 1.0229 - accuracy: 0.5783 - val_loss: 0.9732 - val_accuracy: 0.5882
Epoch 91/250
40/40 [==============================] - 0s 7ms/step - loss: 1.0289 - accuracy: 0.5668 - val_loss: 0.9912 - val_accuracy: 0.6054
Epoch 92/250
40/40 [==============================] - 0s 6ms/step - loss: 1.0213 - accuracy: 0.5731 - val_loss: 1.0258 - val_accuracy: 0.5882
Epoch 93/250
40/40 [==============================] - 0s 7ms/step - loss: 0.9944 - accuracy: 0.5920 - val_loss: 0.9707 - val_accuracy: 0.6054
Epoch 94/250
40/40 [==============================] - 0s 7ms/step - loss: 0.9993 - accuracy: 0.5720 - val_loss: 0.9682 - val_accuracy: 0.5931
Epoch 95/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9918 - accuracy: 0.5815 - val_loss: 0.9813 - val_accuracy: 0.5931
Epoch 96/250
40/40 [==============================] - 0s 7ms/step - loss: 1.0184 - accuracy: 0.5552 - val_loss: 1.0086 - val_accuracy: 0.6029
Epoch 97/250
40/40 [==============================] - 0s 6ms/step - loss: 1.0037 - accuracy: 0.5689 - val_loss: 0.9745 - val_accuracy: 0.5907
Epoch 98/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9996 - accuracy: 0.5773 - val_loss: 0.9852 - val_accuracy: 0.5907
Epoch 99/250
40/40 [==============================] - 0s 7ms/step - loss: 1.0144 - accuracy: 0.5289 - val_loss: 0.9736 - val_accuracy: 0.5980
Epoch 100/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9839 - accuracy: 0.5752 - val_loss: 0.9712 - val_accuracy: 0.6005
Epoch 101/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9988 - accuracy: 0.5762 - val_loss: 0.9707 - val_accuracy: 0.6103
Epoch 102/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9882 - accuracy: 0.5857 - val_loss: 0.9955 - val_accuracy: 0.5882
Epoch 103/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0046 - accuracy: 0.5615 - val_loss: 0.9719 - val_accuracy: 0.6029
Epoch 104/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9954 - accuracy: 0.5710 - val_loss: 1.0038 - val_accuracy: 0.5956
Epoch 105/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0047 - accuracy: 0.5762 - val_loss: 0.9877 - val_accuracy: 0.6005
Epoch 106/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0070 - accuracy: 0.5710 - val_loss: 0.9946 - val_accuracy: 0.5882
Epoch 107/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9971 - accuracy: 0.5836 - val_loss: 0.9997 - val_accuracy: 0.5956
Epoch 108/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9900 - accuracy: 0.5573 - val_loss: 0.9944 - val_accuracy: 0.6029
Epoch 109/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0100 - accuracy: 0.5731 - val_loss: 1.0461 - val_accuracy: 0.5858
Epoch 110/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0016 - accuracy: 0.5689 - val_loss: 1.0009 - val_accuracy: 0.5833
Epoch 111/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0066 - accuracy: 0.5563 - val_loss: 0.9630 - val_accuracy: 0.5882
Epoch 112/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0146 - accuracy: 0.5699 - val_loss: 0.9644 - val_accuracy: 0.5980
Epoch 113/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0108 - accuracy: 0.5699 - val_loss: 0.9663 - val_accuracy: 0.5931
Epoch 114/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9927 - accuracy: 0.5815 - val_loss: 0.9727 - val_accuracy: 0.5931
Epoch 115/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0066 - accuracy: 0.5636 - val_loss: 0.9696 - val_accuracy: 0.5931
Epoch 116/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0016 - accuracy: 0.5594 - val_loss: 0.9764 - val_accuracy: 0.5931
Epoch 117/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0027 - accuracy: 0.5510 - val_loss: 0.9689 - val_accuracy: 0.5980
Epoch 118/250
40/40 [==============================] - 0s 6ms/step - loss: 1.0049 - accuracy: 0.5741 - val_loss: 0.9799 - val_accuracy: 0.6029
Epoch 119/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0071 - accuracy: 0.5647 - val_loss: 0.9818 - val_accuracy: 0.5882
Epoch 120/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9799 - accuracy: 0.5794 - val_loss: 0.9792 - val_accuracy: 0.6029
Epoch 121/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9922 - accuracy: 0.5794 - val_loss: 0.9872 - val_accuracy: 0.6054
Epoch 122/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0063 - accuracy: 0.5563 - val_loss: 0.9793 - val_accuracy: 0.5980
Epoch 123/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0038 - accuracy: 0.5668 - val_loss: 0.9997 - val_accuracy: 0.6005
Epoch 124/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0029 - accuracy: 0.5804 - val_loss: 0.9794 - val_accuracy: 0.5931
Epoch 125/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9826 - accuracy: 0.5678 - val_loss: 0.9922 - val_accuracy: 0.5980
Epoch 126/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0078 - accuracy: 0.5699 - val_loss: 1.0110 - val_accuracy: 0.6127
Epoch 127/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9932 - accuracy: 0.5563 - val_loss: 0.9849 - val_accuracy: 0.6029
Epoch 128/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9612 - accuracy: 0.5825 - val_loss: 0.9831 - val_accuracy: 0.6005
Epoch 129/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9720 - accuracy: 0.5878 - val_loss: 1.0389 - val_accuracy: 0.6029
Epoch 130/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0008 - accuracy: 0.5510 - val_loss: 1.0219 - val_accuracy: 0.6054
Epoch 131/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9951 - accuracy: 0.5647 - val_loss: 0.9662 - val_accuracy: 0.6078
Epoch 132/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9945 - accuracy: 0.5794 - val_loss: 0.9605 - val_accuracy: 0.5980
Epoch 133/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0132 - accuracy: 0.5563 - val_loss: 0.9775 - val_accuracy: 0.6078
Epoch 134/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9776 - accuracy: 0.5868 - val_loss: 0.9766 - val_accuracy: 0.6054
Epoch 135/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0047 - accuracy: 0.5804 - val_loss: 0.9947 - val_accuracy: 0.6054
Epoch 136/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9968 - accuracy: 0.5647 - val_loss: 0.9870 - val_accuracy: 0.6054
Epoch 137/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0101 - accuracy: 0.5710 - val_loss: 0.9888 - val_accuracy: 0.5956
Epoch 138/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9861 - accuracy: 0.5836 - val_loss: 0.9937 - val_accuracy: 0.6054
Epoch 139/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9972 - accuracy: 0.5720 - val_loss: 0.9798 - val_accuracy: 0.6127
Epoch 140/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9799 - accuracy: 0.5720 - val_loss: 0.9880 - val_accuracy: 0.5858
Epoch 141/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9929 - accuracy: 0.5710 - val_loss: 0.9849 - val_accuracy: 0.6078
Epoch 142/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9774 - accuracy: 0.5647 - val_loss: 0.9796 - val_accuracy: 0.6029
Epoch 143/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9823 - accuracy: 0.5804 - val_loss: 0.9800 - val_accuracy: 0.5956
Epoch 144/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9992 - accuracy: 0.5710 - val_loss: 0.9833 - val_accuracy: 0.5956
Epoch 145/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0175 - accuracy: 0.5699 - val_loss: 1.0071 - val_accuracy: 0.6078
Epoch 146/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0000 - accuracy: 0.5647 - val_loss: 0.9908 - val_accuracy: 0.6005
Epoch 147/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9804 - accuracy: 0.5752 - val_loss: 0.9774 - val_accuracy: 0.5833
Epoch 148/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9764 - accuracy: 0.5647 - val_loss: 0.9854 - val_accuracy: 0.6078
Epoch 149/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9918 - accuracy: 0.5741 - val_loss: 0.9675 - val_accuracy: 0.5956
Epoch 150/250
40/40 [==============================] - 0s 7ms/step - loss: 0.9910 - accuracy: 0.5752 - val_loss: 0.9648 - val_accuracy: 0.5931
Epoch 151/250
40/40 [==============================] - 0s 7ms/step - loss: 0.9973 - accuracy: 0.5678 - val_loss: 0.9644 - val_accuracy: 0.6078
Epoch 152/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0126 - accuracy: 0.5457 - val_loss: 0.9682 - val_accuracy: 0.6103
Epoch 153/250
40/40 [==============================] - 0s 7ms/step - loss: 1.0003 - accuracy: 0.5689 - val_loss: 0.9617 - val_accuracy: 0.6152
Epoch 154/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9892 - accuracy: 0.5804 - val_loss: 0.9799 - val_accuracy: 0.5931
Epoch 155/250
40/40 [==============================] - 0s 7ms/step - loss: 0.9846 - accuracy: 0.5920 - val_loss: 0.9738 - val_accuracy: 0.6029
Epoch 156/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9964 - accuracy: 0.5794 - val_loss: 0.9653 - val_accuracy: 0.6054
Epoch 157/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9764 - accuracy: 0.5668 - val_loss: 0.9747 - val_accuracy: 0.6078
Epoch 158/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9784 - accuracy: 0.5668 - val_loss: 0.9804 - val_accuracy: 0.6103
Epoch 159/250
40/40 [==============================] - 0s 7ms/step - loss: 0.9970 - accuracy: 0.5699 - val_loss: 0.9807 - val_accuracy: 0.6005
Epoch 160/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9896 - accuracy: 0.5783 - val_loss: 1.0180 - val_accuracy: 0.6054
Epoch 161/250
40/40 [==============================] - 0s 7ms/step - loss: 0.9793 - accuracy: 0.5804 - val_loss: 0.9709 - val_accuracy: 0.6103
Epoch 162/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9763 - accuracy: 0.5762 - val_loss: 0.9752 - val_accuracy: 0.6029
Epoch 163/250
40/40 [==============================] - 0s 7ms/step - loss: 0.9931 - accuracy: 0.5573 - val_loss: 0.9718 - val_accuracy: 0.6054
Epoch 164/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9815 - accuracy: 0.5741 - val_loss: 1.0033 - val_accuracy: 0.5931
Epoch 165/250
40/40 [==============================] - 0s 7ms/step - loss: 0.9936 - accuracy: 0.5741 - val_loss: 0.9972 - val_accuracy: 0.6005
Epoch 166/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9852 - accuracy: 0.5815 - val_loss: 1.0041 - val_accuracy: 0.5931
Epoch 167/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9840 - accuracy: 0.5531 - val_loss: 0.9774 - val_accuracy: 0.5980
Epoch 168/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9865 - accuracy: 0.5773 - val_loss: 0.9831 - val_accuracy: 0.6005
Epoch 169/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0039 - accuracy: 0.5615 - val_loss: 1.0163 - val_accuracy: 0.6029
Epoch 170/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9856 - accuracy: 0.5605 - val_loss: 0.9748 - val_accuracy: 0.6103
Epoch 171/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9865 - accuracy: 0.5752 - val_loss: 0.9969 - val_accuracy: 0.5833
Epoch 172/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9745 - accuracy: 0.5804 - val_loss: 0.9940 - val_accuracy: 0.6054
Epoch 173/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9695 - accuracy: 0.5836 - val_loss: 0.9727 - val_accuracy: 0.6127
Epoch 174/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9725 - accuracy: 0.5983 - val_loss: 1.0018 - val_accuracy: 0.6127
Epoch 175/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9677 - accuracy: 0.5899 - val_loss: 0.9989 - val_accuracy: 0.5980
Epoch 176/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9873 - accuracy: 0.5647 - val_loss: 0.9943 - val_accuracy: 0.6054
Epoch 177/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9526 - accuracy: 0.5657 - val_loss: 1.0038 - val_accuracy: 0.5956
Epoch 178/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9881 - accuracy: 0.5741 - val_loss: 0.9740 - val_accuracy: 0.5907
Epoch 179/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9984 - accuracy: 0.5857 - val_loss: 0.9804 - val_accuracy: 0.5882
Epoch 180/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9658 - accuracy: 0.5868 - val_loss: 0.9856 - val_accuracy: 0.5980
Epoch 181/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9708 - accuracy: 0.5783 - val_loss: 0.9884 - val_accuracy: 0.5882
Epoch 182/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9953 - accuracy: 0.5636 - val_loss: 1.0905 - val_accuracy: 0.6054
Epoch 183/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9920 - accuracy: 0.5699 - val_loss: 0.9996 - val_accuracy: 0.5956
Epoch 184/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9832 - accuracy: 0.5752 - val_loss: 0.9685 - val_accuracy: 0.6103
Epoch 185/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9709 - accuracy: 0.5689 - val_loss: 0.9637 - val_accuracy: 0.5931
Epoch 186/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9622 - accuracy: 0.5752 - val_loss: 0.9495 - val_accuracy: 0.6029
Epoch 187/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9736 - accuracy: 0.5699 - val_loss: 0.9487 - val_accuracy: 0.6103
Epoch 188/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9779 - accuracy: 0.5699 - val_loss: 0.9579 - val_accuracy: 0.5980
Epoch 189/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9955 - accuracy: 0.5584 - val_loss: 0.9983 - val_accuracy: 0.5907
Epoch 190/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9893 - accuracy: 0.5783 - val_loss: 1.0072 - val_accuracy: 0.5907
Epoch 191/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9852 - accuracy: 0.5731 - val_loss: 0.9861 - val_accuracy: 0.5956
Epoch 192/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9684 - accuracy: 0.5868 - val_loss: 0.9867 - val_accuracy: 0.5956
Epoch 193/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9792 - accuracy: 0.5741 - val_loss: 1.0155 - val_accuracy: 0.6078
Epoch 194/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9942 - accuracy: 0.5731 - val_loss: 0.9809 - val_accuracy: 0.5980
Epoch 195/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9858 - accuracy: 0.5762 - val_loss: 0.9669 - val_accuracy: 0.6078
Epoch 196/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9761 - accuracy: 0.5868 - val_loss: 0.9837 - val_accuracy: 0.5956
Epoch 197/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9900 - accuracy: 0.5762 - val_loss: 0.9771 - val_accuracy: 0.5980
Epoch 198/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9719 - accuracy: 0.5678 - val_loss: 0.9661 - val_accuracy: 0.5980
Epoch 199/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9639 - accuracy: 0.5794 - val_loss: 0.9709 - val_accuracy: 0.6005
Epoch 200/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9937 - accuracy: 0.5773 - val_loss: 0.9834 - val_accuracy: 0.6029
Epoch 201/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9759 - accuracy: 0.5752 - val_loss: 1.0000 - val_accuracy: 0.6005
Epoch 202/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9615 - accuracy: 0.5647 - val_loss: 0.9750 - val_accuracy: 0.6054
Epoch 203/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9775 - accuracy: 0.5699 - val_loss: 0.9540 - val_accuracy: 0.6078
Epoch 204/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9757 - accuracy: 0.5920 - val_loss: 0.9677 - val_accuracy: 0.6078
Epoch 205/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9729 - accuracy: 0.5741 - val_loss: 0.9670 - val_accuracy: 0.6005
Epoch 206/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9712 - accuracy: 0.5794 - val_loss: 0.9994 - val_accuracy: 0.5858
Epoch 207/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9860 - accuracy: 0.5636 - val_loss: 1.0025 - val_accuracy: 0.5931
Epoch 208/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9897 - accuracy: 0.5773 - val_loss: 1.0010 - val_accuracy: 0.5980
Epoch 209/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9909 - accuracy: 0.5689 - val_loss: 0.9903 - val_accuracy: 0.5833
Epoch 210/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9759 - accuracy: 0.5794 - val_loss: 1.0030 - val_accuracy: 0.5760
Epoch 211/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9503 - accuracy: 0.5868 - val_loss: 0.9683 - val_accuracy: 0.6005
Epoch 212/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9922 - accuracy: 0.5762 - val_loss: 0.9700 - val_accuracy: 0.6078
Epoch 213/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9779 - accuracy: 0.5783 - val_loss: 0.9673 - val_accuracy: 0.5833
Epoch 214/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9557 - accuracy: 0.5825 - val_loss: 0.9646 - val_accuracy: 0.6005
Epoch 215/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9439 - accuracy: 0.5952 - val_loss: 0.9679 - val_accuracy: 0.5980
Epoch 216/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9839 - accuracy: 0.5678 - val_loss: 0.9806 - val_accuracy: 0.6005
Epoch 217/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9652 - accuracy: 0.5899 - val_loss: 0.9604 - val_accuracy: 0.6103
Epoch 218/250
40/40 [==============================] - 0s 7ms/step - loss: 0.9939 - accuracy: 0.5615 - val_loss: 0.9749 - val_accuracy: 0.6078
Epoch 219/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9517 - accuracy: 0.5889 - val_loss: 0.9702 - val_accuracy: 0.6029
Epoch 220/250
40/40 [==============================] - 0s 7ms/step - loss: 0.9725 - accuracy: 0.5762 - val_loss: 0.9807 - val_accuracy: 0.6054
Epoch 221/250
40/40 [==============================] - 0s 7ms/step - loss: 0.9840 - accuracy: 0.5794 - val_loss: 0.9710 - val_accuracy: 0.6201
Epoch 222/250
40/40 [==============================] - 0s 7ms/step - loss: 0.9555 - accuracy: 0.5773 - val_loss: 0.9774 - val_accuracy: 0.6029
Epoch 223/250
40/40 [==============================] - 0s 7ms/step - loss: 0.9735 - accuracy: 0.5825 - val_loss: 0.9879 - val_accuracy: 0.6054
Epoch 224/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9779 - accuracy: 0.5668 - val_loss: 1.0050 - val_accuracy: 0.5931
Epoch 225/250
40/40 [==============================] - 0s 7ms/step - loss: 0.9779 - accuracy: 0.5741 - val_loss: 0.9747 - val_accuracy: 0.6005
Epoch 226/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9906 - accuracy: 0.5825 - val_loss: 0.9869 - val_accuracy: 0.6127
Epoch 227/250
40/40 [==============================] - 0s 7ms/step - loss: 0.9919 - accuracy: 0.5573 - val_loss: 1.0006 - val_accuracy: 0.6054
Epoch 228/250
40/40 [==============================] - 0s 7ms/step - loss: 0.9585 - accuracy: 0.5878 - val_loss: 1.0134 - val_accuracy: 0.5956
Epoch 229/250
40/40 [==============================] - 0s 8ms/step - loss: 0.9810 - accuracy: 0.5699 - val_loss: 0.9926 - val_accuracy: 0.5931
Epoch 230/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9790 - accuracy: 0.5752 - val_loss: 0.9436 - val_accuracy: 0.6054
Epoch 231/250
40/40 [==============================] - 0s 7ms/step - loss: 0.9791 - accuracy: 0.5689 - val_loss: 0.9480 - val_accuracy: 0.5931
Epoch 232/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9646 - accuracy: 0.5846 - val_loss: 0.9613 - val_accuracy: 0.6005
Epoch 233/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9763 - accuracy: 0.5773 - val_loss: 0.9562 - val_accuracy: 0.5907
Epoch 234/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9456 - accuracy: 0.6078 - val_loss: 0.9881 - val_accuracy: 0.6005
Epoch 235/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9637 - accuracy: 0.5868 - val_loss: 0.9548 - val_accuracy: 0.5907
Epoch 236/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9683 - accuracy: 0.5783 - val_loss: 0.9951 - val_accuracy: 0.6005
Epoch 237/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9811 - accuracy: 0.5647 - val_loss: 0.9704 - val_accuracy: 0.6103
Epoch 238/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9662 - accuracy: 0.5794 - val_loss: 0.9770 - val_accuracy: 0.5931
Epoch 239/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9748 - accuracy: 0.5720 - val_loss: 1.0041 - val_accuracy: 0.6054
Epoch 240/250
40/40 [==============================] - 0s 5ms/step - loss: 1.0033 - accuracy: 0.5573 - val_loss: 1.0230 - val_accuracy: 0.5882
Epoch 241/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9731 - accuracy: 0.5699 - val_loss: 0.9723 - val_accuracy: 0.6029
Epoch 242/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9656 - accuracy: 0.5594 - val_loss: 0.9781 - val_accuracy: 0.6029
Epoch 243/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9538 - accuracy: 0.5952 - val_loss: 0.9697 - val_accuracy: 0.5980
Epoch 244/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9684 - accuracy: 0.5846 - val_loss: 1.0026 - val_accuracy: 0.6176
Epoch 245/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9739 - accuracy: 0.5720 - val_loss: 0.9639 - val_accuracy: 0.5956
Epoch 246/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9678 - accuracy: 0.5605 - val_loss: 0.9750 - val_accuracy: 0.6127
Epoch 247/250
40/40 [==============================] - 0s 6ms/step - loss: 0.9781 - accuracy: 0.5499 - val_loss: 0.9658 - val_accuracy: 0.6005
Epoch 248/250
40/40 [==============================] - 0s 5ms/step - loss: 0.9684 - accuracy: 0.5783 - val_loss: 0.9732 - val_accuracy: 0.6029
Epoch 249/250
40/40 [==============================] - 0s 4ms/step - loss: 1.0039 - accuracy: 0.5584 - val_loss: 1.0509 - val_accuracy: 0.5956
Epoch 250/250
40/40 [==============================] - 0s 4ms/step - loss: 0.9714 - accuracy: 0.5857 - val_loss: 0.9779 - val_accuracy: 0.6103

E. Plot visuals as in Q3.C and share insights about difference observed in both the models¶

In [225]:
# Plot the training and validation loss
plt.plot(history_v2.history['loss'])
plt.plot(history_v2.history['val_loss'])
plt.title('Training Loss vs. Validation Loss')
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.legend(['Training Loss', 'Validation Loss'])
plt.show()
In [226]:
# Plot the training and validation accuracy
plt.plot(history_v2.history['accuracy'])
plt.plot(history_v2.history['val_accuracy'])
plt.title('Training Accuracy vs. Validation Accuracy')
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.legend(['Training Accuracy', 'Validation Accuracy'])
plt.show()

In the first model(sd_cls_model), 2 dense layers with tanh activation & sigmoid was used in output layer. Adam was used as optimizer. The accuracy was ranging about 50%

In the next model(sd_cls_model_v1), 4 dense layers were used with relu activation function. Softmax was used as activation function in the output layer. Due to introduction of droput, noise was induced in the layer. SGD was used as optimizer. However, the accuracy has mrginaly improved from previous result.

In the final model(sd_cls_model_v2), 3 dense layers were used with relu as activation function. Output layer used sigmoid activation fucntion. Adam optimizer was used. Number of epochs increased from 100 to 250. This provide better accuracy of about 60% compared to the aboe models.

PART B¶

1. Data Import and Exploration¶

A. Read the .h5 file and assign to a variable

In [2]:
import io
import h5py

path = '/content/drive/My Drive/AIML Course/Introduction to Neural Networks/'
In [3]:
# Open the file as readonly
h5f = h5py.File(path + 'Autonomous_Vehicles_SVHN_single_grey1.h5', 'r')

B. Print all the keys from the .h5 file

In [33]:
h5f.keys()
Out[33]:
<KeysViewHDF5 ['X_test', 'X_train', 'X_val', 'y_test', 'y_train', 'y_val']>

C. Split the data into X_train, X_test, Y_train, Y_test

In [34]:
X_train = h5f['X_train'][:]
y_train = h5f['y_train'][:]
X_test = h5f['X_test'][:]
y_test = h5f['y_test'][:]

2. Data Visualisation and preprocessing¶

A. Print shape of all the 4 data split into x, y, train, test to verify if x & y is in sync

In [35]:
print('X_train' ,X_train.shape)
print('y_train' ,y_train.shape)
print('X_test' ,X_test.shape)
print('y_test' ,y_test.shape)

X_train (42000, 32, 32)
y_train (42000,)
X_test (18000, 32, 32)
y_test (18000,)

B. Visualise first 10 images in train data and print its corresponding labels

In [7]:
%matplotlib inline
import matplotlib.pyplot as plt

columns=10
rows=10

fig=plt.figure(figsize=(12, 8))

for i in range(1,columns*rows+1):
  img=X_train[i]
  fig.add_subplot(rows,columns,i)
  print(y_train[i],end='   ')
  if i % columns == 0:
    print ("")
  plt.imshow(img,cmap='gray')

plt.show()

6   7   4   4   0   3   0   7   3   1   
0   1   3   1   1   0   0   8   4   6   
5   7   9   1   0   3   0   7   2   1   
1   0   2   9   0   2   5   1   3   2   
7   9   8   4   9   4   5   9   3   4   
0   5   5   8   3   6   6   0   0   6   
8   8   3   1   4   7   0   2   9   4   
8   7   3   9   4   4   3   3   4   4   
8   6   8   7   4   0   4   4   4   1   
4   2   7   4   2   9   1   9   1   0

C. Reshape all the images with appropriate shape update the data in same variable

In [36]:
X_train.shape
Out[36]:
(42000, 32, 32)
In [37]:
#Reshaping train and test data from 3d to 2d numpy array
X_train = X_train.reshape((X_train.shape[0], -1))
X_test = X_test.reshape((X_test.shape[0], -1))
In [16]:
X_test.shape
Out[16]:
(18000, 1024)
In [38]:
print("X_train",X_train.shape," X_test", X_test.shape)

X_train (42000, 1024)  X_test (18000, 1024)

D. Normalise the images i.e. Normalise the pixel values

In [39]:
# converting the images from ..input/>.h5 file to numpy array
data = np.array(h5f)
test_x = np.array(h5f['X_test'])
train_x = np.array(h5f['X_train'])
test_y = np.array(h5f['y_test'])
train_y = np.array(h5f['y_train'])
In [40]:
# Cheking type of converted array
print(type(test_x))
print(type(train_x))
print(type(test_y))
print(type(train_y))

<class 'numpy.ndarray'>
<class 'numpy.ndarray'>
<class 'numpy.ndarray'>
<class 'numpy.ndarray'>
In [41]:
# use min-max scaler to normalise the pixel values (make max value as 1 and min value as 0)
scale_minmax=MinMaxScaler()

train_x=scale_minmax.fit_transform(train_x.reshape(-1,train_x.shape[-1])).reshape(train_x.shape)
test_x=scale_minmax.transform(test_x.reshape(-1,test_x.shape[-1])).reshape(test_x.shape)
In [42]:
# Checking Lowest and Highest Pixel Value after normalization of images
print("Lowest Pixel Value:", train_x[0].min(),"Highest Pixel Value:", train_x[0].max())

Lowest Pixel Value: 0.049835496 Highest Pixel Value: 0.51063794

2.E. Transform Labels into format acceptable by Neural Network

In [43]:
# converting y data into categorical (one-hot encoding)
y_train = to_categorical(y_train)
y_test = to_categorical(y_test)
In [44]:
print('y_train',y_train.shape, 'y_test',y_test.shape)

y_train (42000, 10) y_test (18000, 10)

2.F. Print total Number of classes in the Dataset

In [47]:
num_classes = len(np.unique(train_y))

print("Number of classes:", num_classes)

Number of classes: 10
In [46]:
train_y.shape
Out[46]:
(42000,)

3. Model Training & Evaluation using Neural Network¶

A. Design a Neural Network to train a classifier

In [79]:
train_x_flat = train_x.reshape(X_train.shape[0], -1)
test_x_flat = test_x.reshape(X_test.shape[0], -1)
In [81]:
test_x_flat.shape
Out[81]:
(18000, 1024)
In [85]:
train_x_flat.shape
Out[85]:
(42000, 1024)
In [96]:
model_cls = Sequential()
model_cls.add(Dense(32, input_shape=(1024,)))                  
model_cls.add(Activation('relu'))    
model_cls.add(Dense(16))                   
model_cls.add(Activation('relu'))    
model_cls.add(Dense(16))                   
model_cls.add(Activation('relu'))    
model_cls.add(Dense(16))                    
model_cls.add(Activation('relu'))    
model_cls.add(Dense(10))
model_cls.add(Activation('softmax'))
    
sgd = optimizers.SGD(learning_rate = 0.001)
model_cls.compile(optimizer = sgd, loss = 'categorical_crossentropy', metrics = ['accuracy'])

B. Train the classifier using previously designed Architecture (Use best suitable parameters).

In [97]:
hist_cl_hi5=model_cls.fit(train_x_flat,y_train,batch_size=20,epochs=50,verbose=0,validation_data=(test_x_flat,y_test))

C. Evaluate performance of the model with appropriate metrics.

In [98]:
result_cls = model_cls.evaluate(test_x_flat,y_test)

563/563 [==============================] - 1s 2ms/step - loss: 1.0585 - accuracy: 0.6723
In [99]:
result_cls_train_acc = model_cls.evaluate(train_x_flat,y_train)

1313/1313 [==============================] - 2s 2ms/step - loss: 1.0150 - accuracy: 0.6833

D. Plot the training loss, validation loss vs number of epochs and training accuracy, validation accuracy vs number of epochs plot and write your observations on the same

In [106]:
# Plot the training and validation loss
plt.plot(hist_cl_hi5.history['loss'])
plt.plot(hist_cl_hi5.history['val_loss'])
plt.title('Training Loss vs. Validation Loss')
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.legend(['Training Loss', 'Validation Loss'])
plt.show()
In [107]:
# Plot the training and validation accuracy
plt.plot(hist_cl_hi5.history['accuracy'])
plt.plot(hist_cl_hi5.history['val_accuracy'])
plt.title('Training Accuracy vs. Validation Accuracy')
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.legend(['Training Accuracy', 'Validation Accuracy'])
plt.show()
In [108]:
hist_cl_hi5_v1=model_cls.fit(train_x_flat,y_train,batch_size=10,epochs=50,verbose=0,validation_data=(test_x_flat,y_test))
In [109]:
result_cls_v1 = model_cls.evaluate(test_x_flat,y_test)

563/563 [==============================] - 2s 4ms/step - loss: 0.8143 - accuracy: 0.7663
In [110]:
result_cls_train_acc_v1 = model_cls.evaluate(train_x_flat,y_train)

1313/1313 [==============================] - 4s 3ms/step - loss: 0.7142 - accuracy: 0.7878
In [111]:
# Plot the training and validation loss
plt.plot(hist_cl_hi5_v1.history['loss'])
plt.plot(hist_cl_hi5_v1.history['val_loss'])
plt.title('Training Loss vs. Validation Loss')
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.legend(['Training Loss', 'Validation Loss'])
plt.show()
In [112]:
# Plot the training and validation accuracy
plt.plot(hist_cl_hi5_v1.history['accuracy'])
plt.plot(hist_cl_hi5_v1.history['val_accuracy'])
plt.title('Training Accuracy vs. Validation Accuracy')
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.legend(['Training Accuracy', 'Validation Accuracy'])
plt.show()

The Model was created with 1 Input layer, 2 dense layers & 1 output later. Dense layer has actiavtion function as relu, output function used softmax as activation function:

  1. In the first model training (hist_cl_hi5), batch size used was 20 with 50 epochs. Accuracy achieved was 67% which is at par with training accuracy of 68%

  2. In the second model training (hist_cl_hi5_v1), batch size was reduced to 10 with 50 epochs. The accuracy has improved to about 76% which is close to training accuracy of 78%. However, this training result has more noise compared ot the earlier one. The earlier trained model is comparitively better.

In [ ]: